AI Eats The World, And Most Of Its Flash Storage
… If you do the math, that’s 8.5 exabytes of internal flash capacity and 16.5 exabytes of networked flash capacity, for a total of 25 exabytes of capacity, for a 1 gigawatt installation. …
… If you do the math, that’s 8.5 exabytes of internal flash capacity and 16.5 exabytes of networked flash capacity, for a total of 25 exabytes of capacity, for a 1 gigawatt installation. …
… Micron no doubt does not want to do a lot of such SCA deals because it limits its pricing levers. But it probably is a good thing to ink a few such deals so it can confidently build out a capacity expansion plan and fund it. …
… What is more important, I think, is that MLPerf does not provide either pricing or power consumption information, which is also part of pricing. …
… The real spending levels are what matters because that tells you about capacity, not price or revenue. But of course, a dollar also gets you a lot more capacity than it used to, in general, and this chart does not take that into account. …
… If AMD was doing a bin sort, you would expect a distribution of compute cores and HBM memory capacity being sold, not just one configuration. But the capacity of the MI350P is precisely half of what the MI350X offers. …
… For now, Anthropic is renting TPU capacity on Google Cloud, and that is definitely not the cheapest way to get AI compute engines. …
… If you wanted to get better balance on HBM capacity and bandwidth against the inherent compute in a socket, you might want to actually gear down the number of GPU chiplets, run them at slightly higher clocks, and ramp up the amount and speed of the HBM memory to get more capacity and bandwidth per … …
… The overall Microsoft Cloud business only grew at 39 percent in the quarter – a little lower than was expected – and even though Hood told Wall Street that Microsoft was using a lot of its GPU capacity for other things – presumably to train its own models, to run inference for copilots and other ap… …
… The second category is a capacity requirement. …
… HBM memory capacity, which is vital for most AI accelerators, has probably been largely allocated, although a case could be made that Micron Technology, Samsung, and SK Hynix have very little incentive to lock down HBM pricing now for future sales because, just like DRAM and flash, every time anyon… …